Why are vowels distinguishable merely by formant frequencies and not formant intensities?

Question

So it is commonly said that what distinguishes the qualities of different vowels in a language are the frequencies of the first two or three peaks (formants) on a spectogram, so that one can plot all vowels in a two- or three-dimensional space with very little loss of information.

But I'm a little surprised by this. Even if we barely perceive formants above F5 or so, why is it not the case that the relative intensities of different formants are perceptually important as well? E.g. instead of classifying vowels like "F1 is around 720 Hz, F2 is around 1800 Hz," one could imagine classifying vowels like "F1 is about 720, F2 is around 1800 Hz and has 60% of the intensity of F1, F3 is around 2500 Hz with 20% of the intensity of F1"?

I understand that as an empirical question, just writing down F1-F3 values is good enough to distinguish vowels in all (?) known languages, so I guess I'm looking for some perceptual or physiological reason why we don't need to do some kind of more complicated Fourier analysis that takes their relative intensities into account.

Answer 1

Actually, they are, sort of, but not in English. The underlying reason is that formants are the result of filtering the glottal source wave to enhance resonant frequencies. There is no physical mechanism that adds amplitude to just F1. The general pattern of the glottal source wave is that most of the energy is at the fundamental, and amplitude decreases according to which harmonic you have.

To the extent that you can control the slope of the glottal source (make it flatter; make it fall more steeply), you can control the amplitude of an individual formant (though not to the point of making F3 much higher in amplitude compared to F1). A number of languages (e.g. Dinka, Hmong, Taa) employ a contrastive breathy phonation which has a rapidly falling amplitude in the glottal source wave. Thus, if you can control the source you can weakly control the amplitude of a formant. If you can speak Dinka, you can control the source (or you could just learn how, somewhere).

For amusement, you can synthesize complex waves from individual harmonic and unnaturally modify formant amplitudes (if you can synthezise a complex wave from harmonics), and check the perceptual effect.